Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by retrieving relevant documents from external sources to improve factual accuracy and verifiability. However, this reliance introduces new attack surfaces within the retrieval pipeline, beyond the LLM itself. While prior RAG attacks have exposed such vulnerabilities, they largely rely on manipulating user queries, which is often infeasible in practice due to fixed or protected user inputs. This narrow focus overlooks a more realistic and stealthy vector: instructional prompts, which are widely reused, publicly shared, and rarely audited. Their implicit trust makes them a compelling target for adversaries to manipulate RAG behavior covertly. We introduce a novel attack for Adversarial Instructional Prompt (AIP) that exploits adversarial instructional prompts to manipulate RAG outputs by subtly altering retrieval behavior. By shifting the attack surface to the instructional prompts, AIP reveals how trusted yet seemingly benign interface components can be weaponized to degrade system integrity. The attack is crafted to achieve three goals: (1) naturalness, to evade user detection; (2) utility, to encourage use of prompts; and (3) robustness, to remain effective across diverse query variations. We propose a diverse query generation strategy that simulates realistic linguistic variation in user queries, enabling the discovery of prompts that generalize across paraphrases and rephrasings. Building on this, a genetic algorithm-based joint optimization is developed to evolve adversarial prompts by balancing attack success, clean-task utility, and stealthiness. Experimental results show that AIP achieves up to 95.23% ASR while preserving benign functionality. These findings uncover a critical and previously overlooked vulnerability in RAG systems, emphasizing the need to reassess the shared instructional prompts.

Retrieval Augmented Generation (RAG) frameworks improve the accuracy of large language models (LLMs) by integrating external knowledge from retrieved documents, thereby overcoming the limitations of models' static intrinsic knowledge. However, these systems are susceptible to adversarial attacks that manipulate the retrieval process by introducing documents that are adversarial yet semantically similar to the query. Notably, while these adversarial documents resemble the query, they exhibit weak similarity to benign documents in the retrieval set. Thus, we propose a simple yet effective Graph-based Reranking against Adversarial Document Attacks (GRADA) framework aiming at preserving retrieval quality while significantly reducing the success of adversaries. Our study evaluates the effectiveness of our approach through experiments conducted on five LLMs: GPT-3.5-Turbo, GPT-4o, Llama3.1-8b, Llama3.1-70b, and Qwen2.5-7b. We use three datasets to assess performance, with results from the Natural Questions dataset demonstrating up to an 80% reduction in attack success rates while maintaining minimal loss in accuracy.

Retrieval augmented generation (RAG) systems provide a method for factually grounding the responses of a Large Language Model (LLM) by providing retrieved evidence, or context, as support. Guided by this context, RAG systems can reduce hallucinations and expand the ability of LLMs to accurately answer questions outside the scope of their training data. Unfortunately, this design introduces a critical vulnerability: LLMs may absorb and reproduce misinformation present in retrieved evidence. This problem is magnified if retrieved evidence contains adversarial material explicitly intended to promulgate misinformation. This paper presents a systematic evaluation of RAG robustness in the health domain and examines alignment between model outputs and ground-truth answers. We focus on the health domain due to the potential for harm caused by incorrect responses, as well as the availability of evidence-based ground truth for many common health-related questions. We conduct controlled experiments using common health questions, varying both the type and composition of the retrieved documents (helpful, harmful, and adversarial) as well as the framing of the question by the user (consistent, neutral, and inconsistent). Our findings reveal that adversarial documents substantially degrade alignment, but robustness can be preserved when helpful evidence is also present in the retrieval pool. These findings offer actionable insights for designing safer RAG systems in high-stakes domains by highlighting the need for retrieval safeguards. To enable reproducibility and facilitate future research, all experimental results are publicly available in our github repository. https://github.com/shakibaam/RAG_ROBUSTNESS_EVAL

While neural ranking models (NRMs) have shown high effectiveness, they remain susceptible to adversarial manipulation. In this work, we introduce Few-Shot Adversarial Prompting (FSAP), a novel black-box attack framework that leverages the in-context learning capabilities of Large Language Models (LLMs) to generate high-ranking adversarial documents. Unlike previous approaches that rely on token-level perturbations or manual rewriting of existing documents, FSAP formulates adversarial attacks entirely through few-shot prompting, requiring no gradient access or internal model instrumentation. By conditioning the LLM on a small support set of previously observed harmful examples, FSAP synthesizes grammatically fluent and topically coherent documents that subtly embed false or misleading information and rank competitively against authentic content. We instantiate FSAP in two modes: FSAP-IntraQ, which leverages harmful examples from the same query to enhance topic fidelity, and FSAP-InterQ, which enables broader generalization by transferring adversarial patterns across unrelated queries. Our experiments on the TREC 2020 and 2021 Health Misinformation Tracks, using four diverse neural ranking models, reveal that FSAP-generated documents consistently outrank credible, factually accurate documents. Furthermore, our analysis demonstrates that these adversarial outputs exhibit strong stance alignment and low detectability, posing a realistic and scalable threat to neural retrieval systems. FSAP also effectively generalizes across both proprietary and open-source LLMs.

--In Large Language Models, Retrieval-Augmented Generation (RAG) systems can significantly enhance the performance of large language models by integrating external knowledge. However, RAG also introduces new security risks. Existing research focuses mainly on how poisoning attacks in RAG systems affect model output quality, overlooking their potential to amplify model biases. For example, when querying about domestic violence victims, a compromised RAG system might preferentially retrieve documents depicting women as victims, causing the model to generate outputs that perpetuate gender stereotypes even when the original query is gender neutral. T o show the impact of the bias, this paper proposes a Bias Retrieval and Reward Attack (BRRA) framework, which systematically investigates attack pathways that amplify language model biases through a RAG system manipulation. We design an adversarial document generation method based on multi-objective reward functions, employ subspace projection techniques to manipulate retrieval results, and construct a cyclic feedback mechanism for continuous bias amplification. Experiments on multiple mainstream large language models demonstrate that BRRA attacks can significantly enhance model biases in dimensions. In addition, we explore a dual stage defense mechanism to effectively mitigate the impacts of the attack. This study reveals that poisoning attacks in RAG systems directly amplify model output biases and clarifies the relationship between RAG system security and model fairness. This novel potential attack indicates that we need to keep an eye on the fairness issues of the RAG system. Large Language Models (LLMs) demonstrate exceptional capabilities in natural language understanding and generation [1], [2]. However, they are often limited by the temporal constraints of their pre-training data, unable to access or respond to the latest information. To address this limitation, Retrieval-Augmented Generation (RAG) [3] has emerged as a promising architecture that combines external knowledge bases with generative models, effectively mitigating LLMs' Linlin Wang, Tianqing Zhu, Laiqiao Qin and Wanlei Zhou are with the Faculty of Data Science, City University of Macau, Macao, China (e-mail: linlinwang.cityu@gmail.com; Longxiang Gao is with the Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center, Qilu University of Technology (Shandong Academy of Sciences), Jinan, China, and also with the Shandong Provincial Key Laboratory of Computing Power Internet and Service Computing, Shandong Fundamental Research Center for Computer Science, Jinan, China (e-mail: gaolx@sdas.org)

This paper concerns corpus poisoning attacks in dense information retrieval, where an adversary attempts to compromise the ranking performance of a search algorithm by injecting a small number of maliciously generated documents into the corpus. Our work addresses two limitations in the current literature. First, attacks that perform adversarial gradient-based word substitution search do so in the discrete lexical space, while retrieval itself happens in the continuous embedding space. We thus propose an optimization method that operates in the embedding space directly. Specifically, we train a perturbation model with the objective of maintaining the geometric distance between the original and adversarial document embeddings, while also maximizing the token-level dissimilarity between the original and adversarial documents. Second, it is common for related work to have a strong assumption that the adversary has prior knowledge about the queries. In this paper, we focus on a more challenging variant of the problem where the adversary assumes no prior knowledge about the query distribution (hence, unsupervised). Our core contribution is an adversarial corpus attack that is fast and effective. We present comprehensive experimental results on both in- and out-of-domain datasets, focusing on two related tasks: a top-1 attack and a corpus poisoning attack. We consider attacks under both a white-box and a black-box setting. Notably, our method can generate successful adversarial examples in under two minutes per target document; four times faster compared to the fastest gradient-based word substitution methods in the literature with the same hardware. Furthermore, our adversarial generation method generates text that is more likely to occur under the distribution of natural text (low perplexity), and is therefore more difficult to detect.

Retrieving relevant context is a common approach to reduce hallucinations and enhance answer reliability. Explicitly citing source documents allows users to verify generated responses and increases trust. Prior work largely evaluates citation correctness - whether cited documents support the corresponding statements. But citation correctness alone is insufficient. To establish trust in attributed answers, we must examine both citation correctness and citation faithfulness. In this work, we first disentangle the notions of citation correctness and faithfulness, which have been applied inconsistently in previous studies. Faithfulness ensures that the model's reliance on cited documents is genuine, reflecting actual reference use rather than superficial alignment with prior beliefs, which we call post-rationalization. We design an experiment that reveals the prevalent issue of post-rationalization, which undermines reliable attribution and may result in misplaced trust. Our findings suggest that current attributed answers often lack citation faithfulness (up to 57 percent of the citations), highlighting the need to evaluate correctness and faithfulness for trustworthy attribution in language models.

Recent work showed that retrieval based on embedding similarity (e.g., for retrieval-augmented generation) is vulnerable to poisoning: an adversary can craft malicious documents that are retrieved in response to broad classes of queries. We demonstrate that previous, HotFlip-based techniques produce documents that are very easy to detect using perplexity filtering. Even if generation is constrained to produce low-perplexity text, the resulting documents are recognized as unnatural by LLMs and can be automatically filtered from the retrieval corpus. We design, implement, and evaluate a new controlled generation technique that combines an adversarial objective (embedding similarity) with a "naturalness" objective based on soft scores computed using an open-source, surrogate LLM. The resulting adversarial documents (1) cannot be automatically detected using perplexity filtering and/or other LLMs, except at the cost of significant false positives in the retrieval corpus, yet (2) achieve similar poisoning efficacy to easilydetectable documents generated using HotFlip, and (3) are significantly more effective than prior methods for energy-guided generation, such as COLD. Many modern retrieval systems use embeddings, i.e., dense vector representations, of documents and queries to enable retrieval based on semantic similarity. Chaudhari et al. (2024) and Zhong et al. (2023) recently demonstrated that an adversary can use HotFlip Ebrahimi et al. (2018) to generate documents whose embeddings have high similarity to, and will thus be retrieved in response to, broad classes of queries. We first demonstrate that adversarial documents produced by HotFlip have much higher perplexity than normal text and can be filtered out with negligible collateral damage (i.e., false positives).

The robustness of recent Large Language Models (LLMs) has become increasingly crucial as their applicability expands across various domains and real-world applications. Retrieval-Augmented Generation (RAG) is a promising solution for addressing the limitations of LLMs, yet existing studies on the robustness of RAG often overlook the interconnected relationships between RAG components or the potential threats prevalent in real-world databases, such as minor textual errors. In this work, we investigate two underexplored aspects when assessing the robustness of RAG: 1) vulnerability to noisy documents through low-level perturbations and 2) a holistic evaluation of RAG robustness. Furthermore, we introduce a novel attack method, the Genetic Attack on RAG (\textit{GARAG}), which targets these aspects. Specifically, GARAG is designed to reveal vulnerabilities within each component and test the overall system functionality against noisy documents. We validate RAG robustness by applying our \textit{GARAG} to standard QA datasets, incorporating diverse retrievers and LLMs. The experimental results show that GARAG consistently achieves high attack success rates. Also, it significantly devastates the performance of each component and their synergy, highlighting the substantial risk that minor textual inaccuracies pose in disrupting RAG systems in the real world.

Neural ranking models (NRMs) have undergone significant development and have become integral components of information retrieval (IR) systems. Unfortunately, recent research has unveiled the vulnerability of NRMs to adversarial document manipulations, potentially exploited by malicious search engine optimization practitioners. While progress in adversarial attack strategies aids in identifying the potential weaknesses of NRMs before their deployment, the defensive measures against such attacks, like the detection of adversarial documents, remain inadequately explored. To mitigate this gap, this paper establishes a benchmark dataset to facilitate the investigation of adversarial ranking defense and introduces two types of detection tasks for adversarial documents. A comprehensive investigation of the performance of several detection baselines is conducted, which involve examining the spamicity, perplexity, and linguistic acceptability, and utilizing supervised classifiers. Experimental results demonstrate that a supervised classifier can effectively mitigate known attacks, but it performs poorly against unseen attacks. Furthermore, such classifier should avoid using query text to prevent learning the classification on relevance, as it might lead to the inadvertent discarding of relevant documents.